audio / clock drift

[Evan Gardner]

I am in the full habit of 3 quick claps or finger snaps at the beginning and end of any sessions video or audio. I have found that gives all the participants a firm "go" marker and a "we made it to the end!" marker. 

I like 3 because then it gives a good visual to look for on the audio strip, and it gives three points to line up or match to a video file and a visual reference if needed. Having just one point is just that much harder.

I also like to practice having the participants (either informants or students or skit actors) pausing and freezing for one second after the quick 3 snap in. This gives us some room to fade.

Thanks for breaking down your solution. I'm sure I'll refer to it someday!


> On Sep 19, 2014, at 3:17 PM, Lindsay Marean <lmarean at bensay.org> wrote:
> 
> Thank you, Doug!
> 
> This didn't exactly work, but your overall approach gave the the solution I've been looking for.
> 
> First, to clarify, all our recordings are digital; even digital devices often run at slightly different speeds, according to the Internet.
> 
> I also had to reverse the ratio (video/audio) and changed the audio speed rather than the video speed, because I want the resulting audio track to be in sync with the actual video in ELAN.
> 
> One minor complication is that my audio and video files aren't exactly the same length; of course I can trim the beginning or end off of either file, but for the total duration ratio to work I will need to trim very precisely.
> 
> The reason it didn't work to use Praat is that I couldn't find any Convert menu.  I'm using Praat version 5.3.85.  I found Convert here and there a few places, but nowhere with a "Lenthen..." option.
> 
> However, the basic idea of figuring out the ratio of length between the two files did work.  Here's what I ended up doing (in case anyone else also struggles with this):
> 
> 1)  Open both files in Audacity (the original audio file and the audio file extracted from video).
> 2)  Find a salient sound event as close to the beginning of both recordings as possible, and a salient sound event as near the end of both.  I'm trying to get into the habit of clapping my hands at the beginning and at the end of a session.
> 3)  Carefully measure the distance between the salient sounds on both recordings.
> 4)  Determine the ratio of audio / video.  Convert to a percent by shifting the decimal 2 places to the left.
> 5)  Under the Effects menu in Audacity, change tempo for the audio track by the indicated percent.  If the original audio file is longer, it needs to be a negative percent.
> 6)  Voila!  Both files should be at the same speed now.  Now just align them, and I'm good to go!
> 
> Thanks, and I hope this helps someone else.
> 
> Lindsay
> 
>> On 9/17/14, 10:20 AM, Doug Whalen wrote:
>>   Dear Lindsay,
>>   It seems likely that your video recorder was not digital; hopefully the audio recorder was.  If so, do the following:  Open the two audio files in Praat.  Get the (exact) ratio of the from-audio wav file to the from-video wav file (dur_from_audio/dur_from_video).  Select the from-video sound object.  From the Convert menu, choose "Lengthen (overlap-add)".  This will use the PSOLA algorithm to change the duration but maintain the original F0.  Put your ratio in the "Factor" field and click OK.  This should create a new sound object with the same duration as the from-audio file.  Check to make sure it all seems right.  
>>   Don't forget to save the new sound object!  It's not a file until you save it.
>>   Hope this works, Doug DhW
>> 
>> 
>>> On Sep 15, 2014, at 4:16 PM, Lindsay Marean <lmarean at bensay.org> wrote:
>>> 
>>> I'm hoping that someone here on ILAT has run into this problem and can suggest a way to deal with it:
>>> 
>>> We're recording fluent speakers talking, with both audio recorders and video recorders.  Recently I recorded a session in which an audio recorder picked up one speaker really well, and a video recorder picked up another speaker really well.  I can use Audacity to combine the audio (one on each stereo channel) into a single WAV file that I can then use with ELAN for transcription.
>>> 
>>> The problem is that the two recorders don't record at exactly the same speed.  In a long session, this difference becomes very noticeable - the two tracks may be perfectly synchronized at the beginning, but they will be out of sync by the end.
>>> 
>>> Here are a few things I've found on the Internet that I think discuss the same issue: http://www.acoustica.com/mixcraft/v4/help/hs1550.htm and http://forum.audacityteam.org/viewtopic.php?f=46&t=75868.
>>> 
>>> I've been trying to use the change tempo function of Audacity to adjust one track, but it's been trial-and-error - change by a small value, see if it works, and then when it doesn't, undo the change, and change by a different small value.  So far my approach is really time-consuming but still not really successful for making a transcribable track.
>>> 
>>> Has anyone else dealt with this problem?  How?  Does anyone know of a better way to get both tracks moving at the same speed, beginning to end?
>>> 
>>> Thank you!
>>> Lindsay Marean
>> 
>> 
>> 
>> Douglas H. Whalen, President 
>> Endangered Language Fund 
>> 300 George St., Suite 900 
>> New Haven, CT 06511 
>> USA 
>> +1-203-865-6163, ext. 265 (or 234 for Whalen) 
>> elf at endangeredlanguagefund.org 
>>  www.endangeredlanguagefund.org
> 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.linguistlist.org/pipermail/ilat/attachments/20140919/544c24ea/attachment.htm>